At one extreme are users whose need for word processing is satisfied by Google Docs. At the other extreme are layout experts for whom Scribus is the only feasible app.
In-between are those like publishers and technical writers who produce long, text-oriented documents. This category of users is served by Adobe FrameMaker on Windows, and to some extent by LibreOffice Writer on Linux.
Unfortunately, these users are apparently not a priority in LibreOffice, Calligra Words, AbiWord, or any other office suite. Features that would provide for these users include:
Browser extensions have taught me how useful color coded tabs can be for workspaces. The titles of open tabs disappear when more than eight or nine or open, so the color is often the quickest visual guide to the relation between tabs.
The same system could be just as useful on the desktop. Better yet, the color coding might be preserved between sessions, allowing users to open all the apps needed for a specific task at the same time. So far, I know of no desktop with such a feature.
For years, Stardock Systems has been selling a Windows extension called Fences, which lets icons be grouped. You can name each group and move the icons in it together. In addition, you can assign which fence different types of files are automatically added to, and hide and arrange fences as needed.
In other words, fences automate the sort of arrangements that users make on their desktop all the time. Yet aside from one or two minor functions they share with KDE's Folder Views, fences remain completely unknown on Linux desktops. Perhaps the reason is that designers are focused on mobile devices as the source of ideas, and fences are decidedly a feature of the traditional workstation desktop.
As I made this list, what struck me was how few of the improvements were general. Several of these improvement would appeal largely to specific audiences, and only one even implies the porting of a proprietary application. At least one is cosmetic rather than functional.
What this observation suggests is that, for the general user, Linux has very little left to add. As an all-purpose desktop, Linux arrive some years ago, and has been diversifying ever since, until today users can choose from over half a dozen major desktops.
None of that means, of course, that specialists wouldn't have other suggestions. In addition, changing needs can make improvements desirable that nobody once cared about. But it does mean that many items on a list of desirable improvements will be highly personal.
All of which raises the question: what other improvements do you think would benefit the desktop?